NTT Research Unveils Physics of AI Group to Enhance AI Understanding

NTT Research has launched the Physics of Artificial Intelligence Group, a new initiative aimed at advancing the understanding and trust of AI systems. This group, which emerged from the Physics & Informatics (PHI) Lab, will be led by Dr. Hidenori Tanaka, an expert in physics, neuroscience, and machine learning. The group aims to address the 'black box' nature of AI, focusing on trustworthiness and safety in AI applications, announced in a press release.

The Physics of Artificial Intelligence Group will continue collaborations with academic institutions such as Harvard University and Princeton University. The group's mission includes deepening the understanding of AI mechanisms, integrating ethics into AI systems, and improving human-AI collaboration. This interdisciplinary approach combines physics, neuroscience, and psychology to support goals like fairness and safety in AI adoption.

The group builds on previous contributions, including a neural network pruning algorithm and a bias-removal algorithm for large language models. It also plans to explore energy-efficient AI computing platforms, inspired by the efficiency of biological brains compared to artificial neural networks.